Cache vs. Clock vs. RAM

From: James Noyes (jnoyes-sml@retrogeeks.com)
Date: Fri Feb 04 2005 - 12:45:41 EST


Fellow managers:
   I've recently spec'ed a hardware acquisition for a database-and-java-heavy
inventory management application to be deployed on at my workplace. Our
current system is running on a 2x250MHz E3000, if that gives you any hints.
:) I spec'ed a 2x1200 V480, which seemed FAR more than adequate for our
needs.
   The vendor is now asking why I didn't spec a "newer, faster, better,
cheaper" V440. Looking at Sun's data, I can see that basically what we'd get
in the change is a small jump to a 2x1280 clock speed, and a change to SDRAM
versus DRAM(?) in the 480. The FC-AL to U320 change won't likely have much
impact much since the disk mechanisms themselves will be the storage
bottleneck, not the bus. But most importantly, we'd also lose MOST of the L2
cache, with only 1M on the 440 (US-IIIi) versus 8M on the 480 (US-III).
   My personal experience tells me that L2 cache makes a fairly significant
difference in the overall performance of a system, but I don't really have
anything other than anecdotal evidence and "gut feelings" to back this up.
What do the rest of you managers think/feel about this? Is the small clock
speed increase and faster RAM going to be enough to make up for the
substantially smaller cache? Is the fact that this application is primarily
java and database instead of, say, web serving using C or perl or whatever
going to affect how it reacts to the smaller cache?
   Additional anecdotal evidence and personal experience is welcomed, but I'd
love to also have some hard data to present as well. Unfortunately, I'm a bit
under the gun, since they asked the question this morning, and we were
supposed to finalize the spec by close of business today. I look forward to
any input you might offer, and I'll summarize when it's all over.

Thanks in advance,
James Noyes
(jnoyes-sml@retrogeeks.com)
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers



This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:30:07 EDT